Hadoop job throwing ConnectException accessing port 8021

Question

I've installed CDH5 using Cloudera Manager. I copied and built the word count application from a Cloudera Map/Reduce tutorial

When I run it I get a ConnectException with the following text:

Exception in thread "main" java.net.ConnectException: Call From fatman.localdomain/10.1.1.10 to fatman.localdomain:8021 failed on connection exception: java.net.ConnectException: Connection refused; For more details see: http://wiki.apache.org/hadoop/ConnectionRefused

I'm new to MRv2 and Yarn so I did a lot of research trying to figure out what is going on here. My research indicates that Port 8021 is usually associated with the JobTracker in MRv1 and that there is no JobTracker in Yarn. Based on what I found at link in the exception message, there seems to be three possible problems, a firewall is dropping the connection, there is a configuration error or a service that should be listening on that port is not running. I have verified that the firewalls are all turned off. I suspect that there is no service listening to port 8021 and that there shouldn't be one and this is a configuration problem, but I don't know how to verify that. I have installed and uninstalled several different CDH versions prior to arriving at my current state. Something could have gotten messed up in that process.

I have located a hadoop configuration file /etc/hadoop/conf/mapred-site.xml with the following properties:

<property>
    <name>mapred.job.tracker</name>
    <value>fatman.localdomain:8021</value>
</property>
<property>
    <name>mapred.job.tracker.http.address</name>
    <value>0.0.0.0:50030</value>
</property>

So I could be getting the 8021 port from this configuration file. When I looked into this I discovered that CDH5 has MRv1 ports defined including 8021. So this leads me to believe CDH5 supports both MRv1 and MRv2. When I look in the Cloudera Manager I see a "Yarn (MR2 Included)" service running and nothing that looks like it would be associated with MRv1. This leads me to believe I'm running MRv2 and not MRv1.

So I'm stumped. I suspect the word count application should not be trying use port 8021 in MRv2 (can anyone confirm this?). Does anyone know what is going on here or have any ideas on how to figure out what is going on?

Thanks, Steve

UPDATE: I have resolved this problem. I assumed it was a configuration problem and carefully un-installed the CDH making sure to delete everything associated with the installation and then re-installed. Everything worked fine after that. So...I didn't really figure out exactly what was wrong, but I am able to move forward now.

Thank you Mr. Chowdary for your answer. I would up vote it but I am lacking reputation to do so.

Mr.Chowdary Mr.Chowdary · Accepted Answer · 2015-01-09T13:56:37

It is quite typical to find the issue without having configuration files and complete error stacks.
In MR2, No longer we use JobTracker, so configuring of mapred-site.xml with jobtracker is not using by hadoop. yarn-site.xml is used for resource manager configuration and enabling yarn model resource management and application monitoring.
So you must define in yarn-site.xml where is resource manager is running like below

 <property>
    <name>yarn.resourcemanager.address</name>
    <value>host:port</value>
    <description>the host is the hostname of the ResourceManager and the port is the port on
    which the clients can talk to the Resource Manager. </description>
  </property>

And check whether SSH is configured, without configuring this, Application Master can't communicate with node managers to launch containers.
You can refer the excellent sites this and this for easily configuring YARN in single node cluster
Hope it helps!

Hadoop job throwing ConnectException accessing port 8021

1 Answers